Photo: Torc

A couple weeks back, self-driving tech company Torc Robotics was testing one of its cars in public when a driver in a Kia almost ran it off the road—purposely. The driver was “super aggressive in front of us,” said Michael Fleming, Torc’s CEO, “and then they switched back into the middle lane, and we both came to a stoplight.”

The Kia driver then rolled down his window, got the attention of Torc’s employees inside the startup’s car, and—according to Fleming—said, “Hey, does that thing really drive itself?”

“We said, ‘Yes, it does,’” Fleming continued. “He said, ‘Well, I intentionally cut you off to see what it’d do.”

That’s the kind of real-world scenarios Torc has found itself in while testing autonomous car technology over the last decade. The kind that need to be planned for if the industry wants to succeed with its long-shot bid to make autonomous cars a reality. And while it’s hardly a household name like Google or Uber or any car company, Torc is actually quite ahead of much of the pack despite its tiny size.

At the Consumer Electronics Show in Las Vegas this week, self-driving cars are front and center, thanks in part to recent investments from massive companies like Intel and Toyota. But Torc, a relatively tiny startup compared to some big players in the industry, has shown you don’t need massive infusions of cash to produce super-capable autonomous cars.

Advertisement

Photo: Torc

Founded in 2005, the startup’s based in Blacksburg, Virginia and has only 80 employees on staff right now. Despite the smaller presence, Torc has managed to develop cars with just as much capability—if not more—than some of the biggest players in the field right now: their cars, dubbed Asimov (for the science fiction writer), can perform U-turns, drive seamlessly in inclement weather, perform high-speed merges (sometimes across several lanes of traffic), and obey rules at three-way stops.

Fleming thinks the company’s advanced to the point it has and stands out from the super-crowded AV field thanks to his staff’s immense experience, both with its employees and extensive testing in the field. The company’s employees have been kicking around the self-driving car space for more than a decade. And Torc’s aggressive about gathering as much input from the public and its own data from trips.

Advertisement

At CES, Torc offered rides to consumers who were picked at random from respondents of a public survey conducted at the company’s booth. Torc’s cars have been tested in 20 states. Last summer, the company sent Asimov on a cross-country jaunt, which netted a boatload of information to absorb.

“Our cross country trip exposed us to a lot of different environments,” Fleming told Jalopnik in an interview. “Well marked roads, roads that aren’t marked well.” Las Vegas also has unique quirks to handle: the city’s street layout is terrible and, instead of clean painted lines, Vegas uses Botts’ dots.

CES proved it’s got a huge step on the field. While, unfortunately, I wasn’t able to get a test drive in Torc’s vehicle this week, I caught a glimpse of Asimov in action on Tuesday, when numerous autonomous car demo rides were canceled by companies at CES due to rain.

Advertisement

Weather clearly wasn’t a deterrent for Torc; coming out of dinner that night, I saw one of its vehicle pass me by on the street like it was no big deal.

Earlier in the week, one of Torc’s cars found itself in a particularly sketchy situation. After successfully pulling a U-turn, Asimov came across a white truck heading in the opposite direction toward it.

Without any intervention—like almost everyone, Torc keeps a safety engineer behind the driver’s seat just in case the technology flubs—the car perceived the wrong-way driver and safely came to a stop.

Some drivers may have just scooted around the truck to its left, but Fleming was quick to point out that Asimov had a vehicle to its side that also stopped for the vehicle. What’s striking is that Torc’s car doesn’t stutter. It’s driving smoothly, like a smart human driver would.

Advertisement

“These are those crazy corner cases that we only encounter every once in awhile,” Fleming said .”That’s what we’re addressing.”

Fleming goes way back in the autonomous car space. A decade ago, he participated in the DARPA Grand Challenges, the well-known events that really brought autonomous cars into the limelight for the first time. So he’s not some advantageous Silicon Valley darling looking to jump on the bandwagon, he’s a true believer in this stuff.

Like everyone else, in the first DARPA challenge, Fleming’s team didn’t finish the race. But in the final run in 2007, Torc developed the software stack for Virginia Tech’s team, which was one of three groups to actually cross the finish line. If there was ever an indication he had a solid group, this was it. While the DARPA events led to self-driving car teams being scooped up by the likes of Google and others, Fleming managed to keep his Virginia-centric squad together.

Advertisement

“What I noticed was, we had such a talented group of students and some great self driving technology ... and I wanted to keep a winning team intact, andI wanted to see us commercialize this technology,” Fleming said.

A decade later, Torc’s generating serious buzz across the industry. It’s one of the few remaining startups that has yet to be purchased by a major company—surely there’s some who’d be interested in purchasing and securing its expertise—and it’s shown you don’t need the backing of a big name to advance the goal of developing autonomous tech.

Though he’s an idealist, Fleming knows the industry has a long way to go to perfect the technology. He didn’t say whether the company intends to eventually sell or license its tech to a bigger player, but Fleming believes the advancement of autonomous tech requires the sort of robust partnerships that’ve emerged over the last year between Silicon Valley and traditional automakers.

Advertisement

“I don’t think there’s one entity,” he said, “that’s going to bring forth the entire self-driving system.”

Correction: Due to an oversight on my part, a previous version of this post stated that Torc had 400 employees, when the company actually has 80. I regret this error.